Reopening a review you finished shows the result you earned. A completed review that you came back to said "we cannot assess this" for a moment on the conclusion row, in the sidebar and on the results screen, then corrected itself once it had re-checked — and if the re-check failed, it stayed that way and saved the wrong state. It now opens on your result straight away.

A review that hasn't been re-checked says so, and offers to re-check. Reopening a review that was held up by a conflict used to leave you on a screen with nothing on it. It now tells you nothing has been checked since the review last changed, and gives you a button to check again.

The conclusion premise takes an acceptance only. Rejecting a step withholds it from the reckoning — but the conclusion is what the argument set out to establish, so it stays in either way and a rejection there would have changed nothing it appeared to change. The reject control is no longer offered on it, and the review says why.

A value your answers settle both ways is always reported. Where your answers and the steps you accepted make something true and false at once, the results screen names the claims involved — even when every other figure on the screen reads clean, which is exactly the case that was slipping through.